Link Aggregation on IOS-XR
The IOS-XR term for link aggregation is a link bundle. A link bundle can use LACP or a static “on” ether channel in which there is no protection or negotiation mechanism. In a link bundle, multiple interfaces are bundled together and act like a single physical point-to-point link.
You can bundle multiple links from different line cards, giving you line card redundancy. The bandwidth of a bundle is the aggregate of all member interfaces’ bandwidth. So a bundle of two 10gig interfaces has a bandwidth of 20gig from the OS’s perspective.
In IOS-XR you can bundle up to 64 individual links. By default only 8 links will be active but you can change this with the below command:
RP/0/RP0/CPU0:XR1(config)#int be1
RP/0/RP0/CPU0:XR1(config-if)#bundle maximum-active links ?
<1-64> Maximum number of active links in this bundle
RP/0/RP0/CPU0:XR1(config-if)#bundle maximum-active links
If you configure the max-active links as a number less than the total links, then you will have an active/passive setup in which some links are passive and waiting to become active.
The ASR 9000 series supports mixed speed bundles, so you can bundle 1G and 10G interfaces together, but the ratio cannot exceed 10. So you cannot bundle 1G and 100G interfaces together.
You configure a bundle interface as follows:
<member interfaces>
bundle id # mode active|passive|on
You then configure the bundle port as if it was a physical interface:
int BE1
ip address 1.1.1.1/30
When using mode on it is called Etherchannel. It does not use LACP, instead it statically bundles the port unconditionally. If you do not specify the mode, it will default to mode on.
RP/0/RP0/CPU0:XR1(config)#int gi0/0/0/3
RP/0/RP0/CPU0:XR1(config-if)#bundle id 3
RP/0/RP0/CPU0:XR1(config-if)#commit
RP/0/RP0/CPU0:XR1(config-if)#do sho run int gi0/0/0/3
Wed Oct 12 02:28:21.483 UTC
interface GigabitEthernet0/0/0/3
bundle id 3 mode on
By default the LACP timer is the slow timer which is 30 seconds. This means that it can take 3x 30sec = 90 secs to detect an indirect link failure! To change this, use the following command on a member interface:
int Gi0/0/0/0
lacp period short|<100-1000 msec>
This command is a bit counter intuitive, because it tells the peer on this interface what speed to use. So if you leave the other end to the default slow timer, the local system will continue sending at 30 sec intervals, and the peer will send at 1 second intervals. For example, I configured only Gi0/0/0/0 for a short timer in my lab. Notice that only the received LACPDU counter has climbed excessively.
RP/0/RP0/CPU0:XR1#show run int gi0/0/0/0
Wed Oct 12 01:20:01.530 UTC
interface GigabitEthernet0/0/0/0
bundle id 1 mode active
lacp period short
!
RP/0/RP0/CPU0:XR1#show lacp counters
Wed Oct 12 01:20:03.235 UTC
Bundle-Ether1
LACPDUs Timeouts
Port Sent Received Excess Expired Defaulted
-------------- ---------------------------------- ----------------------
Gi0/0/0/0 76 199 0 1 1
Gi0/0/0/1 74 45 0 1 1
You can understand how the short timer works by seeing a pcap of the LACPDU. I’ve set the lacp period short, and the router has set the LACP timeout flag to 1, which indicates a short timeout (1s). When set to 0, it indicates a long timeout (30s). This basically tells the peer “my timeout period is short, so you better start sending quick LACPDUs.” In response the peer immediately starts sending LACPDUs every second.

To view information about the bundle, similar to the show etherchannel sum command on IOS, use the following show command:
RP/0/RP0/CPU0:XR1#show bundle
Wed Oct 12 02:49:45.495 UTC
Bundle-Ether1
Status: Up
Local links <active/standby/configured>: 2 / 0 / 2
Local bandwidth <effective/available>: 2000000 (2000000) kbps
MAC address (source): 0022.6b41.99f4 (Chassis pool)
Inter-chassis link: No
Minimum active links / bandwidth: 1 / 1 kbps
Maximum active links: 24
Wait while timer: 2000 ms
Load balancing:
Link order signaling: Not configured
Hash type: Default
Locality threshold: None
LACP: Operational
Flap suppression timer: Off
Cisco extensions: Disabled
Non-revertive: Disabled
mLACP: Not configured
IPv4 BFD: Not operational
IPv6 BFD: Not configured
Port Device State Port ID B/W, kbps
-------------------- --------------- ----------- -------------- ----------
Gi0/0/0/0 Local Active 0x8000, 0x0001 1000000
Link is Active
Gi0/0/0/1 Local Active 0x8000, 0x0002 1000000
Link is Active
Note that the max active links here is only 24. I believe this is a limitation of the virtualized platform. On a 9000 series I see a max active links value of 64.
Further Reading
Last updated